Rotation and divergence biases between scatterometer wind observations and model fields

Abstract

Many atmospheric processes can be characterized by the rotation and divergence of the surface wind field. An accurate portrayal of these phenomena is very important for numerical weather prediction. This project focused on the global and regional model bias between the rotation and divergence of near-surface ocean wind observations, measured by satellite scatterometers, and numerical model fields of the European Center for Medium-range Weather Forecasting operational model (ECMWF-Ops) for the year 2022. The observations of two different scatterometer types are used over a total of five satellites: ASCAT (MetOp-B and -C) and HSCAT (HY-2B, -2C and -2D). On global and yearly time scales, modelled wind convergence is weaker than conver- gence found in the observations. Also, the bias in the wind rotation is minor, less than 10 % of the mean values. However, on some occasions, regional biases are significant. Turbulent island wakes are one of the features that show a significant bias in wind rotation, with an overestimation in strength for the modelled wake near Madagascar and a slightly different orientation for the modelled wakes of Hawaii, which are most likely caused by smoothing of the island topography in ECMWF-Ops. The gap wind in South Mexico, the Tehuano Wind, is well represented in ECMWF- Ops, both in the wind divergence and rotation, resulting in no significant biases. Yet, the gap winds near Hawaii show a positive bias in the divergence fields as well as the wind vector field, likely a result of the much smaller scale of the gaps with respect to the Tehuano Wind. The most significant bias in the wind rotation is located over the Gulf Stream in the Western North Atlantic. Here, ASCAT wind observations show a dual curl-band over the Gulf Stream, with negative rotations on the west-side and positive rotations on the east-side. These rotations, caused by the surface winds responding to large sea surface temperature gradients, are not seen in the model fields, resulting in a large bias. This demonstrates that ECMWF-Ops does not fully resolve air-sea coupling at smaller scales. Overall, HSCAT observations correlate better with ECMWF-Ops compared to AS- CAT observations. Correlations are stronger at mid-latitudes and over significant weather phenomena, otherwise they are highly variable. Finally, it is demonstrated that an extra interpolation step in the collocated model data results in a little more smoothing of the fields, but to no significant changes. Simultaneously, the necessity of collocation of the model fields with the observations is demonstrated using the temporal mean of the complete ECMWF-Ops dataset.